Skip to main content
POST
Rerank documents

Authorizations

Authorization
string
header
required

Your Seekr API key, sent in the Authorization header with no 'Bearer' prefix.

Body

application/json
query
required

Query to score the documents against.

documents
required

Documents to score against the query.

model
string | null

ID of the model to use.

user
string | null

Identifier for the end user, accepted for compatibility. Ignored by vLLM; other backends may use it.

truncate_prompt_tokens
integer | null

Maximum number of prompt tokens to keep. Use -1 for the model context limit. truncation_side selects which end is dropped; when unset, the backend uses its default.

Required range: x >= -1
truncation_side
enum<string> | null

Which side to truncate from when truncate_prompt_tokens is active. 'right' keeps the first N tokens. 'left' keeps the last N tokens.

Available options:
left,
right
request_id
string | null

The request_id related to this request. If the caller does not set it, a random uuid will be generated.

priority
integer
default:0

The priority of the request (lower means earlier handling; default: 0). Any priority other than 0 will raise an error if the served model does not use priority scheduling.

mm_processor_kwargs
Mm Processor Kwargs · object | null

Additional kwargs to pass to the HF processor.

cache_salt
string | null

If specified, the prefix cache will be salted with the provided string to prevent an attacker from guessing prompts in multi-user environments.

use_activation
boolean | null

Whether to use activation for the pooler outputs. null uses the pooler's default, which is true in most cases.

max_tokens_per_query
integer
default:0

Maximum number of tokens per query. Queries longer than this will be truncated. 0 means no query-level truncation is applied.

max_tokens_per_doc
integer
default:0

Maximum number of tokens per document. Documents longer than this will be truncated. 0 means no document-level truncation is applied (only truncate_prompt_tokens applies to the combined query+document).

instruction
string | null

Task instruction prepended to each scored pair via the chat template. Equivalent to passing chat_template_kwargs={'instruction': ...}.

chat_template_kwargs
Chat Template Kwargs · object | null

Additional keyword args to pass to the chat template renderer. Will be accessible by the score/rerank chat template.

top_n
integer
default:0

Number of top-scoring documents to return. Set to 0 to return all documents.

Required range: x >= 0

Response

Successful response.

id
string
required

ID of the response.

model
string
required

ID of the model that generated the response.

usage
RerankUsage · object
required
results
RerankResult · object[]
required

Documents ranked from most to least relevant. Each result includes its index in the original list.

Last modified on July 31, 2026